rcarmo/piclaw

Registry

Summary

Piclaw is a self-hosted AI coding agent built with TypeScript and Bun, designed to operate in isolated Docker workspaces with VNC access. It leverages LLMs and adaptive cards to provide an interactive, browser-based coding environment.

Key features

• Self-hosted AI coding agent with Bun runtime
• Isolated Docker workspaces for secure execution
• VNC access for visual interaction with the environment
• Adaptive cards for rich, interactive UI
• Web-based interface for easy management
• Supports multiple LLM backends

Use cases

• Automated code generation and review in isolated environments
• Interactive coding assistance with visual feedback via VNC
• Self-hosted AI agent for private or sensitive projects
• Educational tool for learning programming with AI guidance
• Rapid prototyping and experimentation in disposable workspaces

README excerpt

# `piclaw` — your self-hosted AI workspace  PiClaw packages the [Pi Coding Agent](https://github.com/badlogic/pi-mono) into a self-hosted workspace with a streaming web UI, persistent state, multi-provider LLM support, and a practical built-in toolset that includes [many add-ons](https://rcarmo.github.io/piclaw-addons/). It is for people who want one stateful agent workspace they can run locally or in a container without stitching together half a dozen separate services. ## Why PiClaw  - **One workspace, one app** — chat, editor, terminal, viewers, boards, uploads, and automation in the same web UI - **Persistent state** — SQLite-backed messages, media, tasks, token usage, encrypted keychain, and session-scoped SSH / Proxmox / Portainer profiles - **Practical built-ins** — code editing, Office/PDF/CSV/image/video viewing, draw.io, VNC, browser automation, image processing, MCP, infra tools, and optional cross-instance IPC for paired remote peers - **Agent-first workflows** — steering, queued follow-ups, side prompts, autoresearch loops, scheduled tasks, and visual artifact generation - **Context conservation** — small always-active tool baseline with staged discovery via `list_tools` / `list_scripts` - **Optional auth/channels** — passkeys/TOTP for the web UI, plus optional WhatsApp integration ## Quick start ```bash mkdir -p ./home ./workspace docker run -d \ --init \ --name piclaw \ --restart unless-stopped \ -p 8080:8080 \ -e PICLAW_WEB_PORT=8080 \ -v "$(pwd)/home:/config" \ -v "$(pwd)/workspace:/workspace" \ ghcr.io/rcarmo/piclaw:latest ``` Open `http://localhost:8080` and type `/login` to configure your LLM provider, including custom OpenAI-compatible endpoints when you are not us

Topics